Bureaucrats must cooperate to implement public policies. This means working with colleagues from different public organizations, lawyers and accountants from oversight agencies, political appointees, elected officials, etc. Each of these actors holds different, oftentimes conflicting, organizational identities. These attachments go beyond their rational interests and, as I demonstrate in this dissertation, become comparable to other social identities, like race, partisanship, and gender. This means that while bureaucrats see their colleagues as in-groups who share similar values, those working at different organizations are their out-groups who will most likely see the world through different lenses. Relying on a measurement that is well-established in social psychology, I demonstrate that the social distance between different public sector actors helps to understand conflict and cooperation in implementation processes. I refer to this phenomenon as bureaucratic polarization and show that it can change public policies and organizations. The empirical evidence combines face-to-face interviews conducted in two Brazilian states and multiple surveys and experiments fielded in the United States, the United Kingdom, and Brazil. These multiple pieces of research validate the generalizability of bureaucratic polarization as a theoretical framework and an estimation strategy to better understand coordination in public administration and the politics of policy implementation.